From: Jef Allbright (jef@jefallbright.net)
Date: Thu Sep 08 2005 - 19:33:48 MDT
I've been reading and observing this discussion, and note that this may
be the most fertile thread recently in terms of serious thought on all
sides. But the key point has been lost in the haze of warfare over
implementation details and personal competencies. No one has responded
by demonstrating an encompassing understanding of the concept that was
presented graciously and with the warning that apprehension of the big
picture may be obscured by contests over the details. No one has shown
that they grasp the point being made within the context presented, and
then, from that basis, that they might possibly point out what they see
wrong with the concept.
There is an arrogance and self-righteousness about this group that is
ironic in its sincere desire to save all humankind. The result is such
that it misguidedly minimizes the search space in a manner very similar
to the particular conceptual block which was intended to be illuminated
in this thread.
At a paradigmatic level, there is a conception of intelligence as the
ability to affect desired change (desired from the POV of the agent)
within a constrained and uncertain environment. Any "real" and
interesting environment is evolving in a complex (deterministic but
unpredictable) way, and necessarily, the agent must be co-evolving if it
is to survive. (I am not referring to genetic evolution here, but
rather the more fundamental ongoing process of synergetic growth,
punctuated, occasional dead-ends, and all.) My point (and I suspect it
relates to the initial point of this thread) is that the very nature of
creative intelligence within a co-evolving environment is such that the
Self can not be invariant, and thus its goal(s) can not be invariant, if
it is to persist and grow.
No math here, and an imperfect linguistic representation, but a concept
like many others that must remain in the realm of philosophy until our
science of complex systems can make more explicit its (possible)
applicability. In the meantime, perhaps a concept worth considering
since it may help us choose where to allocate our limited time and
resources: attempting to create a single AI with the desirable but
improbable property of Friendliness, or development of intelligence
amplification technology across a broad base of humans already rooted in
human morality but lacking the increased awareness to apply it
effectively to our future.
- Jef
http://www.jefallbright.net
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:52 MDT